Cyber security predictions for 2020

Although many have predicted that AI is the so-called silver bullet that will put an end to cyber-crime, once and for all, I have also identified a paradox, which is the surprising lack of focus on the security of AI itself. This, I predict, will change in 2020. AI security is a problem because AI models are still fairly insecure, and therefore vulnerable to attack. An adversarial approach could look like a malicious hacker intervening in the training process of an AI. For example, an AI learning to recognise cats could be tricked into believing that an image of a dog was also a cat: an exploit that could later be leveraged. It is also possible to extract parts of an AI model, leading to intellectual property theft, as well as the ability to craft "adversarial" AI that could manipulate the intended model.